All proteins undergo local structural fluctuations (LSFs) or breathing motions. These motions are likely to be important for function but are poorly understood. LSFs were initially defined by amide hydrogen exchange (HX) experiments as opening events, which expose a small number of backbone amides to 1H/2H exchange, but whose exchange rates are independent of denaturant concentration. Here, we use size-dependent thiol-disulfide exchange (SX) to characterize LSFs in single cysteine-containing variants of myoglobin (Mb). SX complements HX by providing information on motions that disrupt side chain packing interactions. Most importantly, probe reagents of different sizes and chemical properties can be used to characterize the size of structural opening events and the properties of the open state. We use thiosulfonate reagents (126–274 Da) to survey access to Cys residues, which are buried at specific helical packing interfaces in Mb. In each case, the free energy of opening increases linearly with the radius of gyration of the probe reagent. The slope and the intercept are interpreted to yield information on the size of the opening events that expose the buried thiol groups. The slope parameter varies by over 10-fold among Cys positions tested, suggesting that the sizes of breathing motions vary substantially throughout the protein. Our results provide insight to the longstanding question: how rigid or flexible are proteins in their native states?
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Studying the dynamic characteristics of native proteins remains a challenge in protein chemistry. Dynamic experiments inform us that the native state is not a rigid entity; rather, it is an amalgam of conformations that fluctuate about a mean structure. These fluctuations range from small breathing motions [local structural fluctuations (LSFs)] to whole molecule unfolding. LSFs provide native proteins with dynamic motion, which can be essential for function. For example, binding generally requires some structural rearrangements to accommodate the ligand.1–3 In particular, cooperative binding appears to be facilitated by dynamic motion of the polypeptide chain, which often allows distant binding sites to communicate with one another.4 Finally, local motions have been linked to elementary catalytic steps for a number of enzymes including triosephosphate isomerase,5–7 ribonuclease A,8, 9 and dihydrofolate reductase.10, 11
Amide hydrogen exchange (HX) and thiol-disulfide exchange (SX) are uniquely suited to characterize structural fluctuations in proteins. HX refers to exchange of a hydrogen ion from a backbone amide group with one from the solvent. SX is analogous to HX except that the exchanging species are a Cys side chain and a thiol-reactive reagent added to the solvent. Both processes are modeled by the Linderstrøm–Lang mechanism [Eq. (1)]
where Kop is the equilibrium constant between open and closed states and kint is the intrinsic exchange rate in the absence of structure. The EX2 condition is shown in which open and closed states equilibrate rapidly with respect to kint. In the EX2 limit, the observed exchange rate (kobs) is given by kobs ≅ Kop·kint and the free energy associated with the opening reaction (ΔG or ΔG) is equal to −RT·ln Kop. Exchange takes place according to three classes of structural opening events, in order of increasing size: LSFs, subglobal unfolding, and global unfolding. Subglobal and global unfolding events are identified by an increase in kobs with denaturant concentration. Virtually, all amide and thiol groups exchange by subglobal or global unfolding at sufficiently high-denaturant concentration.12 By contrast, exchange rates of groups exposed by LSFs are independent of denaturant concentration, signifying that little surface area is uncovered in these fluctuations. LSFs tend to be low-energy, high-probability events. As evidence, most backbone hydrogens exchange via LSFs in the absence of denaturant.13, 14 LSFs are a rich source of dynamic information, which has been largely unexplored.
HX has been the main tool used to characterize LSFs to date. Using 2D NMR, HX rates can be measured for many residues simultaneously and in a nonperturbing fashion.13, 15–17 The limitations of HX/NMR are: (i) probe molecules are restricted to hydrogen ion isotopes, (ii) the data are only indicative of opening reactions that expose the backbone amides to exchange, and (iii) resonance assignments are required. With respect to the last point, mass spectrometry detection negates the need for assignment and extends the protein size limit considerably but comes at the expense of residue-specific resolution.
SX complements HX by addressing the limitations described above. The principle advantage of SX is that one can use an extensive selection of thiol-reactive probe reagents to investigate the size and properties of opening reactions. Moreover, SX reports on events that expose buried side chains (rather than backbone amides) to the solvent. Finally, because SX can be monitored by techniques such as absorbance, fluorescence, and mass spectrometry, the protein size limitation is largely eliminated and site-specific resolution is retained. The disadvantage of SX is that Cys-containing mutants must be created and tested for possible perturbation of structure or stability. To facilitate this process, the Harbury group has developed a method for rapidly generating and analyzing large libraries of single-Cys mutants.18
Here, we use thiosulfonate (TS) compounds as probe reagents. TSs react rapidly and specifically with thiolate groups. We selected four compounds ranging in size from the smallest commercially available TS [methyl methanethiosulfonate (MMTS)] to one of the largest ([2-trimethylammonium)ethyl] toluenethiosulfonate, TTS). For the target protein, we use sperm whale myoglobin (Mb) with its heme group in the water-bound ferric state. Wild-type (WT) Mb contains no Cys residues. In a previous study, we created 10 single-Cys mutants and measured their SX rates as a function of urea and guanidine hydrochloride (GdnHCl) concentration using the bulky reagent 5,5′-dithiobis(2-nitrobenzoic acid) (DTNB).19, 20 At low-denaturant concentrations, ΔG values were significantly higher for DTNB than for MMTS at most sites. This disparity suggests that the size difference between reagents can be exploited to estimate the amplitudes of the structural opening events. The purpose of the work presented here is to further investigate the size dependence of SX by using intermediately sized thiol probes to characterize LSFs at helical packing interfaces in Mb.
Mb mutants and thiol probe reagents
We chose to study the V10C, V17C, A90C, and S108C Mb mutants for the following reasons: (i) Their side chains are buried at representative helical packing interfaces: A-H (V10C), A-B-G (V17C), F-H (A90C), and G-H (S108C) [Fig. 1(a)].21 The A-B-G-H helices constitute the stable core of Mb and are known to fold first,22 whereas the F helix folds last and only in the presence of heme.23 (ii) Subsitution with Cys at these positions does not destabilize Mb substantially.20, 21, 24 (iii) These side chains exhibit slow SX rates, consistent with their high degree of burial in the X-ray structure [Fig. 1(b)], yet they are known to exchange via LSFs.20 (iv) Our previous study suggested that ΔG is dependent on probe size at each of these positions.
Figure 2 shows the four TS reagents used in this study: MMTS, [1-(trimethylammonium)methyl] methanethiosulfonate (MTSMT), [2-(trimethylammonium)ethyl] methanthiosulfonate (MTSET), and TTS. Ideally, kint values are obtained using unfolded protein as the substrate (as opposed to short peptides). To do so, we denatured heme-free Mb (apoMb) S108C in 2M GdnHCl. Because TS reagents exchange rapidly with exposed thiols, it was necessary to slow the reaction by lowering pH to 5.0. The second-order intrinsic rate constants (k) were measured for each TS in 2M GdnHCl, pH 5.0 and corrected to match experimental conditions (0M GdnHCl, pH 7.0) by applying the extrapolation terms described in section Materials and Methods. The corrected k values are listed in Table I. kint (which is a pseudo first-order rate constant) was then calculated for each experiment by multiplying the k values in Table I by the TS concentration used in that experiment. We previously determined that kint varies by no more than a factor of two among the Cys mutants tested25; therefore, the kint values obtained for apoMb S108C were used for all variants. Available data indicate that LSFs are rapid and result in EX2-limited exchange.13, 26, 27 As a test, we measured SX rates for Mb A90C using three concentrations of MMTS. Log kobs increased linearly with log·[MMTS] (slope = 0.90, R2 = 0.98; data not shown), verifying the EX2 condition. Sizes of the TS probes were estimated by calculating their radii of gyration (Rg) using SYBYL 6.5.28 These values are summarized in Table I.
Figure 3 shows representative data collected in triplicate for the reaction of Mb S108C with MMTS. Mb is reacted with at least a fivefold excess of TS probe to ensure pseudo first-order kinetics. At the indicated time points, the reaction is quenched by free Cys and desalted. Free protein thiol groups are then labeled with Texas Red (TR) methylthiosulfonate (TRMTS) in denaturing solution. Following a second desalting step to remove unreacted TRMTS, the fraction of free thiol is determined by the molar ratio of protein-bound TR to protein (measured by absorbance). As shown in Figure 3, the data fit to a one-exponential function, which decays approximately to zero. This result is expected and is representative of the majority of data collected. In few select reactions, the signal decayed to a level significantly above zero (not shown). This finding indicates that a fraction of protein was extremely slow to react under native conditions, but reacted normally with TRMTS in 6M GdnHCl. The most likely source of this unreacted material is a minor population of aggregated protein. In these cases, the rate observed for the major, faster-reacting population is reported. Table II summarizes the calculated ΔG values.
Table II. SX Parameters for Ub and Mb Variants
ΔG (kcal mol−1)
ΔG (kcal mol−1)
Errors are standard deviations of at least three measurements.
2.50 ± 0.048
4.08 ± 0.25
3.87 ± 0.88
4.44 ± 0.12
7.58 ± 0.48
8.77 ± 0.21
8.43 ± 0.39
7.78 ± 0.23
5.12 ± 0.26
5.28 ± 0.35
6.10 ± 0.11
6.10 ± 0.26
4.05 ± 0.052
6.87 ± 0.088
6.76 ± 0.13
6.90 ± 0.23
4.24 ± 0.021
5.82 ± 0.004
6.75 ± 0.005
6.98 ± 0.3
SX at a solvent-exposed site
Our goal is to use probes of different sizes to investigate the nature of open states. To establish a benchmark for comparison, we first examine the exchange behavior of a Cys residue at a surface site (position 42) on a folded protein [ubiquitin (Ub)]. The side chain of Arg42 is fully exposed in the X-ray structure of WT Ub [Fig. 1(c)] and exchange of Cys42 in the R42C mutant is expected to occur without a major structural opening event that would require movement of backbone atoms. However, because Cys42 is surrounded by folded structure, it likely mimics the open state of LSFs more realistically than short peptides or denaturant-unfolded proteins.
Figure 4(a) shows the correlation between ΔG and Rg for Ub R42C. One might expect a priori that each probe would exchange at its intrinsic rate. If so, the correlation would follow a flat line with slope (m) and y-intercept (ΔG) equal to zero. Surprisingly, all of the TS probes exchange significantly more slowly (100- to 3400-fold) than kint and the data are minimally fit by a line with m > 0. There are two limiting explanations for why kobs < kint. The first is that Cys42 in native Ub and Cys108 in GdnHCl-denatured apoMb both exchange in an unhindered manner, but the pKa of Cys42 is higher than that of Cys108. Cys42 would thus exchange more slowly than Cys108 because a greater percentage of the Cys42 thiol exists in protonated form. A difference in pKa values would presumably arise from either a high concentration of negative charge surrounding Cys42 or a high concentration of positive charge surrounding Cys108. The electrostatic potential around Cys42 in native Ub is known to be positive29 and the only charged residue close in sequence to Cys108 is Glu109. Most importantly, however, a pKa difference would simply displace the y-intercept of the ΔG versus Rg correlation by a constant increment. In this case, the slope would not deviate from zero, as we observe in our data.
The second and more likely scenario is that the discrepancy between the Ub R42C and apoMb S108C rates is due to differences in the dynamic properties of the open states. The exchange-competent state of Cys42 is the surface of a folded protein in which nearby side chains and backbone groups are constrained in space by native structure. The rugged features of the surrounding landscape may hinder access to the Cys side chain. By contrast, the open state of Cys108 is the GdnHCl-denatured form, which lacks persistent structure and is much more flexible. This view also provides an explanation for the positive slope of ΔG versus Rg [Fig. 4(a)]. Figure 4(a) indicates that on the surface of folded Ub, larger probe reagents are hindered to a greater extent than smaller reagents. This phenomenon is less pronounced or altogether absent in GdnHCl-denatured apoMb. Existing SX and HX studies use either GdnHCl-denatured proteins or short peptides to model the open state. Although this may be appropriate when exchange occurs via global unfolding, an open state such as that of R42C Ub may better reflect the locally unfolded conformations from which LSFs and subglobal exchange takes place.
SX at buried locations
Figure 4(b–e) shows SX results for Mb V10C, V17C, A90C, and S108C, respectively. For most variants, the points fit satisfactorily to a straight line, and we analyze the data using the same m and ΔG parameters as above. This interpretation assumes that the logarithm of the exchange rate depends linearly on Rg values of the probe reagents. However, it is important to recognize that this relationship is purely empirical. To our knowledge, there is no theoretical model from which the dependence of exchange rate on Rg can be derived. We assume linearity because it fits the data using the minimum number of parameters.
Table II summarizes m and ΔG values obtained from the linear fits of the data in Figure 4. A large ΔG value implies an unfavorable event, which may involve breaking a number of interactions. A ΔG value of ≤1 kcal/mol is comparable with that observed for a surface-exposed thiol group and indicates a low-energy fluctuation. It is reasonable to intuit that ΔG is proportional to the physical size of the opening event, although this is not necessarily so. Instead, we propose that the physical size of the fluctuation is reported primarily by m. If the amplitude of the fluctuation is smaller than or comparable with Rg, then smaller probe reagents will react more rapidly than larger molecules. This relationship leads to a positive deflection in the ΔG versus Rg plot, which is manifested by m > 0 in the linear model. If the opening size is larger than the largest TS, then ΔGop will be independent of probe size and m = 0. m for R42C Ub [1.0 kcal/mol/Å; Fig. 4(a)] establishes a reference for the size-dependence of exchange for a Cys residue, which is solvent-exposed but surrounded by native structure. The m values for V10C (0.15 kcal/mol/Å) and V17C (0.63 kcal/mol/Å) are lower than that of the surface Cys control. This result suggests that at these sites the amplitudes of the LSFs are larger than the Rg of the largest probe (3.4 Å). Large fluctuations are expected to be high-energy, low-probability events. Accordingly, ΔG values for these mutants are significantly higher than that of the Ub R42C control (Table II). By contrast, A90C and S108C exhibit larger slopes than the exposed Cys control (m = 1.6 kcal/mol/Å in each case). The small ΔG values (1.9 kcal/mol and 1.7 kcal/mol, respectively) support the notion that fluctuations at these sites are comparable with or smaller than Rg of the probe reagents.
Conventional native-state HX and SX experiments yield two parameters: ΔGop and the dependence of ΔGop on denaturant. By using probe reagents of varying size, this study introduces two additional parameters, m and ΔG, which provide new information on the size of structural opening events. For example, consider Cys17 and Cys90. Both side chains exhibit similar degrees of solvent accessibility (Fig. 1). The smallest probe reagent (MMTS) exchanges with Cys90 (ΔG = 4.0 kcal/mol) more readily than with Cys17 (ΔG = 5.1 kcal/mol). However, the largest TS probe (TTS) exchanges more readily with Cys17 (ΔG = 6.1 kcal/mol) than with Cys90 (ΔG = 6.9 kcal/mol). The reason for the switch is because Cys90 exchanges 125-fold more slowly with TTS than with MMTS, whereas Cys17 is relatively insensitive to probe size. The m value of Cys90 is 2.5-fold greater than that of Cys17. This result suggests that the opening event exposing Cys17 is larger than the opening event exposing Cys90.
Cys90 and Cys108 exhibit large m and small ΔG values, suggesting that the LSFs exposing those side chains occur frequently and are relatively small. Cys90 is located in helix F and Cys108 is found in helix G (Fig. 1). Both side chains pack against side chains in helix H (Ile142 and Leu135/Arg139, respectively). It is possible that minor fluctuations involving a portion of helix H may facilitate exchange at both positions. The side chain of Ser108 is fully buried in the X-ray structure of WT Mb and the side chain of Ala90 is partially accessible (Fig. 1) yet their ΔG and m values are approximately equal. Small-amplitude LSFs can therefore promote exchange at both partially exposed and well-packed positions.
The small m and large ΔG values of Cys10 and Cys17 signify that LSFs at these positions are large and relatively rare. Cys10 and Cys17 are both located in helix A. The Val10 side chain is fully buried and it contacts Ala130 and Met131 at the N-terminus of helix H. The A-H interface thus appears to be quite rigid; if small-amplitude LSFs of helix H facilitate exchange at positions 90 and 108, such fluctuations must be limited to the C-terminal portion of the helix. The side chain of Val17 makes contact with Asp20 and Val21 (helix A), His24 (helix B), and Leu115 (helix G). The packing interfaces between helix A and its neighboring helices thus appear to be quite stable. As noted above, highly buried positions can be exposed by small, low-energy LSFs and the converse is true as well. Cys17 is partially solvent-exposed, yet it requires a large, high-energy LSF to exchange.
It is instructive to compare this work with previous HX and SX studies of Mb. The exchanging species in HX experiments (the hydroxide ion) is smaller than MMTS. Cavagnero et al.30 measured HX rates for CO-bound Mb and these rates were converted to ΔG values according to Eq. (1).26 Val10, Val17, Ala90, and Ser108 can be categorized into two groups based on these HX data. ΔG values are small for Ala90 and Ser108 (4.3 kcal/mol and 4.6 kcal/mol, respectively) and ΔG < ΔG in each case (ΔG = 5.0 kcal/mol for Cys90; ΔG = 5.6 kcal/mol for Cys108), when SX was measured using the bulky DTNB probe.26 This result is consistent with the large m values measured here for Cys90 and Cys108. In contrast, HX of Val10 and Val17 is associated with large ΔG values (8.2 kcal/mol and 7.7 kcal/mol, respectively) and ΔG > ΔG for both residues (ΔG = 7.4 kcal/mol for Cys10; ΔG = 6.5 kcal/mol for Cys17). This finding is in agreement with the notion that positions 10 and 17 are exposed by large-scale, high-energy fluctuations. However, it must be noted that HX and SX data were collected at different pH values, temperatures, and heme ligation states. Therefore, comparisons can be made only on a qualitative basis.
The above discussion considers that each Cys reporter is exposed by an independent fluctuation (with the possible exception of Cys90 and Cys108, which exhibit similar m and ΔG values). In addition, all probe reagents are envisioned to exchange by the same unique fluctuation at any given site. It is also plausible that, at some sites, more than one fluctuation is involved and the probe reagents differentiate between them based on their size. For example, Cys90 appears to exhibit a bimodal distribution of ΔG values [Fig. 4(d)]. This pattern can be explained by two locally unfolded conformations with ΔG values of ∼4 kcal/mol and ∼7 kcal/mol, respectively. The lower energy conformation admits MMTS but not MTSMT, MTSET, or TTS. The higher energy fluctuation admits all probe reagents with approximately equal ease. In this scenario, Cys10 would be accessible only in the higher energy conformation. By observing trends such as these using additional Cys reporters and probe reagents, it may be possible to determine whether LSFs at different sites are independent or are linked by a more limited set of opening events.
Our results address a longstanding paradox posed by HX experiments. LSFs are widely regarded as minor breathing motions which expose little solvent-accessible surface and appear to allow only a single amide NH to exchange at a time.14 Yet LSFs can exhibit ΔGop values that approach ΔG of global unfolding. How can such subtle fluctuations be associated with large opening free energies? Figure 4(a) suggests that part of the answer may be that different kint values apply for global versus subglobal or local unfolding events. HX studies use a common set of reference kint values, which were measured using unstructured model peptides.31 Short peptides represent the globally unfolded state well, as evidenced by the close agreement between ΔGop and ΔGfold for hydrogens that exchange by whole-molecule unfolding. When exchange occurs from LSFs or sub-global unfolding events, however, only a portion of the molecule opens. One can envision that probe molecules may only be able to approach the amide or thiol group by a limited number of trajectories, as opposed to having full 360° access to sites in the globally denatured protein.32 Thus, exchange-competent states in LSFs are likely to resemble the solvent-exposed surface of a structured protein moreso than the globally unfolded state. If so, then kint values may be overestimated for hydrogens that exchange by LSF and subglobal mechanisms. This potential problem, which would result in overestimation of ΔGop for LSFs [by as much as 4.4 kcal/mol in this study; Fig. 4(a)], has been discussed by others in the context of HX.33, 34
In conclusion, we have demonstrated that conformational openings in proteins can be characterized by two new parameters, ΔG and m, which are obtained by size-dependent SX. A freely exchanging Cys on the surface of a folded protein exchanges 100- to 3400-fold more slowly than the intrinsic rates measured with GdnHCl-unfolded protein. Moreover, the former demonstrates an unexpected dependence on probe size that is not exhibited by the latter. Examining four buried positions in Mb reveals that two are exposed to the solvent by small-amplitude fluctuations and two are exposed by larger opening events. The apparent size of the fluctuations is not correlated with extent of side-chain burial, but instead may be dictated by local motions of the helices in Mb.
Materials and Methods
Gene construction and protein purification
The Ub R42C mutation was introduced into the WT Ub gene using the QuikChange kit (Stratagene). Ub and Mb were purified as described.25, 35
Intrinsic rate measurements
All TS reagents were purchased at the highest purity available from Toronto Research Chemicals. ApoMb S108C was reduced with tris(2-carboxyethyl)phosphine (TCEP) for 30 min at room temperature then desalted using a PD-10 column (GE Healthcare) into 2M GdnHCl, 20 mM acetate (pH 5.0). The midpoint of denaturation of apoMb S108C is less than 1M GdnHCl (data not shown). ApoMb was mixed with a fivefold excess of each TS to ensure pseudo first-order conditions. SX was quenched at designated time points by excess cysteine. The protein was then precipitated with 10% trichloroacetic acid and unreacted thiol was measured by DTNB assay. The decay of free thiol was fit to a single-exponential function to obtain the pseudo first-order rate constant for each probe reagent (Supporting Information Fig. S1). k was calculated by dividing the aforementioned rate by the TS concentration. k values reported in Table I are the average of duplicate measurements.
The dependence of thiol reactivity on pH was determined by measuring the exchange rate of MMTS over the pH range 4.5–5.5. The same GdnHCl-unfolded sample of apoMb S108C was used. ApoMb S108C (15 μM) was mixed with MMTS (100 μM) in 2M GdnHCl at pH 4.5, pH 4.75, pH 5.0, and pH 5.5 (0°C). The reaction was quenched and unreacted thiol was measured as above. Plotting log kobs (s−1) versus pH yielded a straight line with a slope of 1.2 (Supporting Information Fig. S2).
The dependence of thiol reactivity on GdnHCl concentration was determined by measuring the rate of DTNB exchange with GdnHCl-unfolded calbindin D9k (I73C variant).36 In each reaction, calbindin (16 μM) was reacted with DTNB (150 μM) in 50 mM Tris (pH 7.5), 20°C. The rate was measured in GdnHCl (4–7M concentration) using a Bio-Logic SFM4-Q/S operating in stopped-flow mode (absorbance detection). Plotting kobs against the concentration of GdnHCl yielded a straight line with a slope of −0.218 s−1M−1 and a y-intercept of 2.02 s−1 (Supporting Information Fig. S3).
Thiol-disulfide exchange experiments
Native Mb was reacted with the thiol probes in the absence of denaturant (pH 7.0, 0°C). Lyophilized Mb was resuspended in reaction buffer (100 mM sodium phosphate, pH 7.0) to a final concentration of 30–40 μM. Concentrations of TS stock solutions were determined at the beginning of each experiment by titration with TNB (ε412 = 14,100M−1 cm−1). The TS concentration in each experiment was adjusted to bring the exchange reaction into roughly the same time course. The final TS concentration exceeded the Mb concentration by at least fivefold in all cases. After mixing TS and Mb, aliquots were removed at various time points, quenched with excess Cys, and immediately precipitated with 10% trichloroacetic acid. Pellets were washed with ice-cold ddH2O and stored at −80°C. To quantitate unreacted thiol, protein pellets were thawed, resuspended in 6M GdnHCl (pH 5.5) and allowed to react fully with excess TRMTS. A Sephadex G25 spin column was used to remove unreacted TRMTS as well as exchange the buffer to 10 mM HCl. TR concentration was determined by absorbance (595 nm) and normalized to Mb concentration, which was determined by measuring sample absorbance at 280 nm and subtracting the calculated TR absorbance at that wavelength. kobs values were obtained by fitting the data to a single exponential decay. Kop values were calculated by dividing kobs by kint, where kint was extrapolated to experimental conditions of temperature, pH, and TS concentration using the data in Supporting Information Figures S2 and S3.
To test for EX2 conditions, SX experiments were repeated as above for Mb A90C at MMTS concentrations of 1.0 mM, 1.5 mM, and 2.0 mM. The observed reaction rate increased with the concentration of probe reagent, as expected for the EX2 mechanism.
The authors thank Tom Duncan for assistance with Rg calculations and Whitney Oswald for technical assistance.